Artificial intelligence’s growing appetite for computing power is shifting the lens through which data center investments are assessed. Access to electricity is no longer the sole bottleneck; the impact on local water resources is becoming a critical factor. A single large facility can consume up to 19 million liters of water daily – equivalent to the needs of a town of 50,000 residents – and in Poland, where water stress is already a concern, this trend demands close attention.
Energy efficiency, measured by the PUE metric, has long dominated environmental evaluations of data centers. But with the rise of AI, a new index is gaining ground: WUE, or water usage effectiveness. Industry reports indicate that 89% of sector representatives now see environmental criteria as more important when choosing a data center, driven not just by ecology but by cost control and regulatory resilience.
In Spain’s Aragon region, the expansion of data centers has sparked public opposition over water pressure. Three planned facilities were approved to use about 756,000 cubic meters of water annually, and the investor has also applied for increased withdrawals at existing sites. In Dublin, both energy and water consumption have fueled resistance to further sector growth.
Poland’s water stress challenge
Poland sits on a sensitive threshold. According to OECD data, the country has around 1,500 cubic meters of renewable freshwater resources per capita per year – below the 1,700 cubic meter mark that defines water stress. That means every new data center project must be evaluated locally, considering available intakes, seasonality, and the capacity of water and sewage infrastructure.
– Based on EU-reported data, data centers account for roughly 0.01% of total annual water consumption in Europe, much less than commonly assumed. However, the key is case-by-case analysis of the specific project, location, and cooling technology used, because water consumption is largely a consequence of design and technological choices made by the investor – said Piotr Kowalski, managing director of the Polish Data Center Association (PLDCA).
For local governments, the responsibility is growing. They need to verify whether a planned facility will overburden water systems, especially during droughts or peak demand periods.
Cooling tech makes the difference
Modern AI servers require rack power densities that have jumped from 8-12 kW a few years ago to 30-80 kW today, and over 100 kW in advanced configurations. This shift is forcing data centers to move from traditional air cooling to liquid cooling and direct-to-chip solutions, which improve heat dissipation but demand more precise management of both energy and water.
– From the ABB Trends 2026-2027 report, AI will become one of the top development priorities for industry within three years. Over one-third of manufacturing companies say so. For the data center sector, this is an important signal because industrial AI deployments will drive even greater demand for data processing and stable computing power – said Tomasz Zieliński, sales director in ABB’s Electrification business in Poland.
He added that data centers are increasingly designed as industrial facilities with real-time parameter analytics. A closed-loop cooling system, like the one ABB recently implemented for a 400 MW AI data center in North Dakota, eliminates water consumption for heat rejection. – This shows that a facility’s impact on local water resources depends not only on its scale but, above all, on the choice of technology and how the infrastructure is designed – Zieliński concluded.
Źródło: WNP.PL, Fot. Hrach Hovhannisyan / Shutterstock
